Overnight 12 hour tape recordings of arterial oxygen saturation (SaO2, pulse oximeter in the beat to beat mode), breathing movements, and airflow were made on 66 preterm infants (median gestational age 34 weeks, range 25-36) who had reached term (37 weeks) and were ready for discharge from the special care baby unit. No infant was given additional inspired oxygen during the study. The median baseline SaO2 was 99.4% (range 88.9-100%). Eight infants had baseline SaO2 values below 97%, the lowest value observed in a study on full term infants. All but one infant had short-lived falls in SaO2 to less than or equal to 80% (desaturations), which were more frequent (5.4 compared with 0.9/hour) and longer (mean duration 1.5 compared with 1.2 seconds) than in full term infants. There was no evidence that gestational age at birth influenced the frequency or duration of desaturations among the preterm infants. The frequency of relatively prolonged episodes of desaturation (SaO2 less than or equal to 80% for greater than or equal to 4 seconds), however, decreased significantly with increasing gestational age (0.5, 0.4, 0.2, and 0.1 episodes/hour in infants at less than or equal to 32, 33-34, 35, and 36 weeks' gestational age, respectively). Analysis of the respiratory patterns associated with such episodes showed that 5% occurred despite both continued breathing movements and continuous airflow. Five infants had outlying recordings: three had baseline SaO2 values of less than 95% (88.9, 92.7, and 93.8%), and two had many prolonged desaturations (14 and 92/hour; median for total group 0.2, 95th centile 2.3). None of these five infants had been considered clinically to have dis order of oxygenation. Although these data are insufficient to provide information about outcome, we conclude that reference data on arterial oxygenation in preterm infants are important to enable the identification of otherwise unrecognized hypoxaemia.